Outcome Prediction in Cancer

This book is organized into 4 sections, each looking at the question of outcome prediction in cancer from a different angle. The first section describes the clinical problem and some of the predicaments that clinicians face in dealing with cancer. Amongst issues discussed in this section are the TNM staging, accepted methods for survival analysis and competing risks. The second section describes the biological and genetic markers and the rôle of bioinformatics. Understanding of the genetic and environmental basis of cancers will help in identifying high-risk populations and developing effective prevention and early detection strategies. The third section provides technical details of mathematical analysis behind survival prediction backed up by examples from various types of cancers. The fourth section describes a number of machine learning methods which have been applied to decision support in cancer. The final section describes how information is shared within the scientific and medical communities and with the general population using information technology and the World Wide Web. * Applications cover 8 types of cancer including brain, eye, mouth, head and neck, breast, lungs, colon and prostate * Include contributions from authors in 5 different disciplines * Provides a valuable educational tool for medical informatics

Head and Neck cancers account for approximately 3.2% of the estimated 1,660,290 new cancer cases for the year 2013 and roughly 1.9% of cancer-related deaths in 2013. In this research, machine learning techniques were employed to predict outcome in cancer patients supporting more objective assessment of the treatments, including surgery, radiation therapy, or chemotherapy. Selection of features capable of distinguishing between the possible outcomes was accomplished by using a highly selective cohort of 61 patients with similar treatment and location of the primary tumor. An accuracy of 80.33% (compared to a baseline majority classifier of 60.66%) was achieved utilizing this cohort. Further, it is shown that this limited cohort has the power to provide valuable information on outcome prediction utilizing as few as four features. Feature selection was drawn from both clinical features and quantitative imaging features including the site of cancer, primary tumor volume, and race.

There have been extensive studies of classification and prediction of cancer outcome with composite gene features that combine functionally related genes together as a single feature to improve the classification and prediction accuracy. Various algorithms have been proposed for feature extraction, feature activity inference, and feature selection, which all claim to improve the prediction accuracy. However, due to the limited test data sets used by each independent study, inconsistent test procedures, and conflicting results, it is difficult to obtain a comprehensive understanding of the relative performances of these algorithms. In this study, various algorithms for the three steps in using composite features for cancer outcome prediction were implemented and an extensive comparison and evaluation were performed by applying testing to seven microarray data sets covering two cancer types and three different clinical outcomes. Also by integrating algorithms in all three different steps, we aimed to investigate how to get the best cancer prediction by using different combinations of these techniques.

"Background. In cancer, large-scale technologies such as next-generation sequencing and microarrays have produced a wide number of genomic features such as DNA copy number alterations (CNA), mRNA expression (EXPR), microRNA expression (MIRNA), and DNA somatic mutations (MUT), among others. Several analyses of a specific type of these genomic data have generated many prognostic biomarkers in many cancer types, and more frequently in breast cancer. However, it is uncertain which of these data is more powerful and whether the best data-type is cancer-type dependent. Objective. Characterize the prognostic power of models obtained from different genomic data types in Breast Cancer (BRCA) from public repositories and to compare the performance of these models with those obtained from data of Mexican patients".

In this paper, the evaluation of two features in predicting the outcomes of patients with bilharziasis bladder cancer has been investigated using an RBF neural network. Prior to prediction, the feature subsets were extracted from the whole set of features for the purpose of providing a high performance of the network. Throughout the analysis of the prognostic feature combinations, two features, histological type and lymph node status, have been identified as the important indicators for outcome prediction of this type of cancer. The highest predictive accuracy reached 85.O% in this study.

We read with interest the paper by Michiels et al on the prediction of cancer with microarrays and the commentary by Ioannidis listing the potential as well as the limitations of this approach (February 5, p 488 and 454). Cancer is a disease characterized by complex, heterogeneous mechanisms and studies to define factors that can direct new drug discovery and use should be encouraged. However, this is easier said than done. Casti teaches that a better understanding does not necessarily extrapolate to better prediction, and that useful prediction is possible without complete understanding (1). To attempt both, explanation and prediction, in a single nonmathematical construct, is a tall order (Figure 1).

"Prognosis after cancer treatment is a constant concern for physicians, patients and their surrounding friends and family. This is one of the reasons that treatment outcomes prediction is such a critical field of research. The sheer magnitude of data generated within a typical radiation oncology clinic each year facilitates the development and eventual validation of predictive and prognostic models. Furthermore, the technological advances driven by data science have enabled the usage of advanced machine learning techniques which can far exceed the performance of previously used conventional techniques.Most cancer patients follow a standard radiation oncology workflow, which among other things includes medical imaging (CT/PET) and the creation of a radiation therapy treatment plan. As these sorts of data are (in theory) present for every patient, they are ideal variables to input into a predictive model. The goal of this thesis was to investigate these two types of pre-treatment input data (diagnostic imaging and dosimetric data) along with patient characteristics to identify associations and create models capable of predicting a cancer patient's treatment response following radiation therapy. The first objective was to investigate dose-volume metrics as predictors of clinical outcomes in a cohort of 422 non-small cell lung cancer (NSCLC) patients who received stereotactic body radiation therapy (SBRT). A correlation between the dose delivered to the region outside the tumor and the occurrence of distant metastasis was revealed. In particular, patients who received above a certain threshold dose were shown to have significantly reduced distant metastasis recurrence rates compared to the rest of the population. This was first shown on 217 patients all of whom were treated with conventional SBRT treatment modalities. Next, a similar analysis was done on 205 patients who were treated with a robotic arm linear accelerator (CyberKnife). It was found that the CyberKnife cohort had both superior distant control and local control, suggesting that under current prescription practices, CyberKnife, as a delivery device, could be superior for treating NSCLC patients with SBRT. The second objective of this thesis was to investigate the usage of a deep learning framework applied to raw medical imaging data in order to predict the overall prognosis of head & neck cancer patients post-radiation therapy. A de novo architecture was built incorporating CT images, resulting in comparable performance to a state-of-the-art study. Furthermore, our model was shown to recognize imaging features (`radiomics') previously shown to be predictive without being explicitly presented with their definition. The final portion of this work was the development of a multi-modal deep learning framework which incorporated CT & PET images along with clinical information. This was compared to the previous architecture built, showing substantial increase in prediction performance for both overall survival and local recurrence. It was also shown to function in the presence of missing data, a common occurrence within the medical landscape.This work demonstrates that pre-treatment prediction of a cancer patient's post-radiation therapy outcomes is possible by learning correlations and building models from readily available data. Future efforts should be put towards data sharing & data curation to enable the creation and validation of models that eventually can be used in the clinic. Ultimately, predictive models should evolve into generative models whereupon one's treatment could be automatically created with the explicit intention of statistically optimizing that patient's outcomes"--

A systematic review of the most current decision models and techniques for disease prevention and treatment Decision Analytics and Optimization in Disease Prevention and Treatment offers a comprehensive resource of the most current decision models and techniques for disease prevention and treatment. With contributions from leading experts in the field, this important resource presents information on the optimization of chronic disease prevention, infectious disease control and prevention, and disease treatment and treatment technology. Designed to be accessible, in each chapter the text presents one decision problem with the related methodology to showcase the vast applicability of operations research tools and techniques in advancing medical decision making. This vital resource features the most recent and effective approaches to the quickly growing field of healthcare decision analytics, which involves cost-effectiveness analysis, stochastic modeling, and computer simulation. Throughout the book, the contributors discuss clinical applications of modeling and optimization techniques to assist medical decision making within complex environments. Accessible and authoritative, Decision Analytics and Optimization in Disease Prevention and Treatment: Presents summaries of the state-of-the-art research that has successfully utilized both decision analytics and optimization tools within healthcare operations research Highlights the optimization of chronic disease prevention, infectious disease control and prevention, and disease treatment and treatment technology Includes contributions by well-known experts from operations researchers to clinical researchers, and from data scientists to public health administrators Offers clarification on common misunderstandings and misnomers while shedding light on new approaches in this growing area Designed for use by academics, practitioners, and researchers, Decision Analytics and Optimization in Disease Prevention and Treatment offers a comprehensive resource for accessing the power of decision analytics and optimization tools within healthcare operations research.

The three-volume set LNCS 9900, 9901, and 9902 constitutes the refereed proceedings of the 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016, held in Athens, Greece, in October 2016. Based on rigorous peer reviews, the program committee carefully selected 228 revised regular papers from 756 submissions for presentation in three volumes. The papers have been organized in the following topical sections: Part I: brain analysis, brain analysis - connectivity; brain analysis - cortical morphology; Alzheimer disease; surgical guidance and tracking; computer aided interventions; ultrasound image analysis; cancer image analysis; Part II: machine learning and feature selection; deep learning in medical imaging; applications of machine learning; segmentation; cell image analysis; Part III: registration and deformation estimation; shape modeling; cardiac and vascular image analysis; image reconstruction; and MR image analysis.

Issues in Cancer Epidemiology and Research / 2011 Edition is a ScholarlyEditions™ eBook that delivers timely, authoritative, and comprehensive information about Cancer Epidemiology and Research. The editors have built Issues in Cancer Epidemiology and Research: 2011 Edition on the vast information databases of ScholarlyNews.™ You can expect the information about Cancer Epidemiology and Research in this eBook to be deeper than what you can access anywhere else, as well as consistently reliable, authoritative, informed, and relevant. The content of Issues in Cancer Epidemiology and Research: 2011 Edition has been produced by the world’s leading scientists, engineers, analysts, research institutions, and companies. All of the content is from peer-reviewed sources, and all of it is written, assembled, and edited by the editors at ScholarlyEditions™ and available exclusively from us. You now have a source you can cite with authority, confidence, and credibility. More information is available at http://www.ScholarlyEditions.com/.

Tumor tissue slides have been used by clinicians to assess cancer patient's condition and indicate prognosis. Several recent studies have suggested that distribution of important immunological biomarkers on tumor tissue slides might help predict survival outcome. These studies rely upon non-parametric Kaplan-Meier survival analysis with Log-rank test to extract statistical insights, which, however, has several disadvantages such as prediction ambiguity and inability to directly model continuous variables. In this study, we engineered 676 features encoding cellular distribution information from multi-spectral tumor tissue images collected from 118 HPV-negative oral squamous cell cancer patients. We leveraged statistical methods and predictive models to explore the predictive power of these features. We identified 18 features as potential survival predictors through Kolmogorov-Smirnov test. Our best model, random forest model, has achieved 58.54% prediction accuracy rate on independent validation dataset. Although the model does not suggest strong predictive power of selected features, evaluation on large scale training data is still needed to further tune model parameters and generate more concrete results.

Advances in high-throughput genomic and proteomic technology have led to a growing interest in cancer biomarkers. These biomarkers can potentially improve the accuracy of cancer subtype prediction and subsequently, the success of therapy. However, identification of statistically and biologically relevant biomarkers from high-throughput data can be unreliable due to the nature of the data--e.g., high technical variability, small sample size, and high dimension size. Due to the lack of available training samples, data-driven machine learning methods are often insufficient without the support of knowledge-based algorithms. We research and investigate the benefits of using knowledge-based algorithms to solve clinical prediction problems. Because we are interested in identifying biomarkers that are also feasible in clinical prediction models, we focus on two analytical components: feature selection and predictive model selection. In addition to data variance, we must also consider the variance of analytical methods. There are many existing feature selection algorithms, each of which may produce different results. Moreover, it is not trivial to identify model parameters that maximize the sensitivity and specificity of clinical prediction. Thus, we introduce a method that uses independently validated biological knowledge to reduce the space of relevant feature selection algorithms and to improve the reliability of clinical predictors. Finally, we implement several functions of this knowledge-based method as a web-based, user-friendly, and standards-compatible software application.